Ensemble TC Track/Genesis Products in Parallel @NCO Resolution Members Daily Frequency Forecast Length NCEP ensemble (AEMN-para) GFS T574L64-33km(12/02/15) 20+1_CTL 00, 06, 12, 18 UTC 16 days (384hrs) CMC ensemble (CEMN) FNMOC ensemble (FEMN) ECMWF ensemble(eemn) GEM L72-66km (02/13/2013) NAVGEM T159L42-80km (02/13/2013) IFS T639/319L62-30/60km 20+1_CTL 00, 12 UTC 16 days (384hrs) 20+1_CTL 00, 12 UTC 16 days (384hrs) 50+1_CTL 00, 12 UTC 15 days (360hrs) Single model ensemble TC track 10-day forecasts and strike probability Single model ensemble TC intensity 10-day forecasts Single model ensemble TC genesis 2-day forecasts Multi-model ensemble TC track 10-day forecasts and strike probability Please visit our website: http://www.emc.ncep.noaa.gov/gmb/jpeng/tc_ens_v1.html
What products are missing? Wish list from NHC, JTWC and CPHC? Single model ensemble TC genesis 5-day forecasts Multi-model ensemble TC genesis 5-day forecasts Single model ensemble TC wind speed probability forecasts Multi-model ensemble TC wind speed probability forecasts???? Challenges on global ensemble TC genesis forecasts The global ensemble (like GEFS) usually predicts weaker vortex than observations. How to define a model vortex as a genesis TC? How to improve the reliability of TC genesis probability forecast? Challenges on global ensemble TC track forecasts Provide the deterministic track forecast and the uncertainty of the forecast How much uncertainty do we need to tell the public? How to make the deterministic track forecast for the super-large spread cycle?
Workshop on Effective Use of Hurricane Ensembles----Panel Discussion Jiayi Peng EMC Ensemble Team November 17, 2015
300 250 GEFS operational TC track forecasts, Atlantic, 2011-2014 GEFS-T254 Implementation: February 14, 2012 Y2011 Y2012 Y2013 Y2014 Track error(nm) 200 150 100 50 0 0 12 24 36 48 72 96 120 Cases Forecast hours 2011 395 351 307 263 227 174 132 89 2012 466 425 385 347 308 246 199 156 2013 211 178 145 118 93 49 22 20 2014 164 144 124 107 91 66 45 30
500 450 400 350 AVNO----GFS-T1534 AEMN----GEFS-T254 PARA ----GEFS-T574 Atlantic, AL03-12, 07/08-11/09, 2015 GEFS-T574 in parallel @NCO AVNO AEMN PARA Track error(nm) 300 250 200 150 100 50 0 0 12 24 36 48 72 96 120 144 168 Forecast hours CASES 194 176 158 141 125 94 68 49 29 17
Ensemble TC Track/Genesis Products in Parallel @NCO Resolution Members Daily Frequency Forecast Length NCEP ensemble (AEMN-para) GFS T574L64-33km(????) 20+1_CTL 00, 06, 12, 18 UTC 16 days (384hrs) CMC ensemble (CEMN) FNMOC ensemble (FEMN) ECMWF ensemble(eemn) GEM L72-66km (02/13/2013) NAVGEM T159L42-80km (02/13/2013) IFS T639/319L62-30/60km 20+1_CTL 00, 12 UTC 16 days (384hrs) 20+1_CTL 00, 12 UTC 16 days (384hrs) 50+1_CTL 00, 12 UTC 15 days (360hrs) Single model ensemble TC track 10-day forecasts and strike probability Single model ensemble TC intensity 10-day forecasts Single model ensemble TC genesis 2-day forecasts Multi-model ensemble TC track 10-day forecasts and strike probability Please visit our website: http://www.emc.ncep.noaa.gov/gmb/jpeng/tc_ens_v1.html
http://www.emc.ncep.noaa.gov/gmb/jpeng/tc_ens_v1.html
What products are missing? Single model ensemble TC genesis 5-day forecasts Multi-model ensemble TC genesis 5-day forecasts Single model ensemble TC wind speed probability forecasts Multi-model ensemble TC wind speed probability forecasts???? Wish list from NHC, JTWC and CPHC?
TC Genesis Target Area: 0-40N, 110E-350E Forecast Time Window: 48-hour genesis NCEP GEFS TC Genesis Probability Forecast Within 48 hours WP97 EP93 AL97 WP97 will form DANAS (WP23, 10/03 06Z) EP93 will form NARDA (EP14, 10/06 18Z) AL97 will form KAREN (AL12, 10/03 06Z)
Verifying (%) 100 90 80 70 60 50 40 30 20 10 0 GEFS Atlantic N=186 2013 TC Genesis Forecast Reliability Diagram 0 10 20 30 40 50 60 70 80 90 100 Forecast (%) 100 90 80 70 60 50 40 30 20 10 0 GEFS E-Pacific N=238 Red line: perfect reliability Green line: distribution of the forecast genesis probability Blue line: the relationship between the forecast and verifying genesis probability 0 10 20 30 40 50 60 70 80 90 100 Forecast (%) Contingent Table OBSERVARTION FORECAST YES NO YES a=22 b=48 NO c=38 d=78 Hit rate(pod)=a/(a+c) 37% Miss rate=c/(a+c) 63% False alarm rate=b/(b+d) 38% Correct rejection rate=d/(b+d) 62% Critical success index=a/(a+b+c) 0.2 OBSERVARTION FORECAST YES NO YES a=42 b=15 NO c=72 d=109 Hit rate(pod)=a/(a+c) 37% Miss rate=c/(a+c) 63% False alarm rate=b/(b+d) 12% Correct rejection rate=d/(b+d) 88% Critical success index=a/(a+b+c) 0.33
Verifying (%) 100 90 80 70 60 50 40 30 20 10 0 GEFS Atlantic N=205 2014 TC Genesis Forecast Reliability Diagram 0 10 20 30 40 50 60 70 80 90 100 Forecast (%) GEFS E-Pacific N=280 Red line: perfect reliability Green line: distribution of the forecast genesis probability Blue line: the relationship between the forecast and verifying genesis probability OBSERVARTION 100 90 80 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100 Forecast (%) OBSERVARTION Contingent Table FORECAST YES NO YES a=31 b=50 NO c=20 d=104 Hit rate(pod)=a/(a+c) 61% Miss rate=c/(a+c) 39% False alarm rate=b/(b+d) 32% Correct rejection rate=d/(b+d) 68% Critical success index=a/(a+b+c) 0.31 FORECAST YES NO YES a=31 b=43 NO c=86 d=120 Hit rate(pod)=a/(a+c) 26% Miss rate=c/(a+c) 74% False alarm rate=b/(b+d) 26% Correct rejection rate=d/(b+d) 74% Critical success index=a/(a+b+c) 0.19
AL04 Debby, June 20-30,2012 Forecast at June 24, 00Z
AL04 Debby, large uncertainty in forecast at June 24, 00Z
TC Genesis Tracker Code Based on Tim Marchok s latest version for genesis track, we make the following changes: (1)Calculation for 300-500hPa temperature anomaly; (2)Calculation for 200-850hPa zonal wind shear; (3)Calculation for 500hPa mean relative humidity; (4)The AL90-99, EP90-99 and WP90-99 are those invest storms named by NHC and JTWC. In case of NHC and JTWC miss potential TC genesis, we create some interested storms based on GFS-T574 operational, Canadian, NOGAPS and ECMWF Deterministic Forecasts, named as HC01, HC02, etc.
Algorithms for TC genesis probabilistic forecast How to define global model TC genesis? The prediction vortices in Global Ensemble Forecast Systems are very weak. (25kts?) Step No.1: (for GEFS, ECMWF, CMC, FNMOC ensembles) We track every vortex by checking: 1)850/700hPa/surface relative vorticity (max) 2)850/700hPa geopotential height (min) 3)Sea level pressure (min) 4)850/700hPa/surface wind speed (min) 5)SLP gradient (0.0015mb/km), Wind speed at 850hPa ( 1.5m/s) 6)Closed SLP contour checked Step No.2: (for GEFS and ECMWF ensemble) We filter those vortices based on the following criteria: 1)Surface maximum wind speed 10kts 2)850hPa maximum vorticity 10**(-4) 1/s 3)300-500hPa temperature anomaly 0.5c
TC Track, Atlantic 2012-2014, GFS(AVNO) VS. GEFS(AEMN) Track error(nm) 250 200 150 100 50 AVNO AEMN Track error(nm) 250 200 150 100 50 AVNO AEMN 2012 2013 0 0 12 24 36 48 72 96 120 Forecast hours Cases 466 425 385 347 308 246 199 156 0 0 12 24 36 48 72 96 120 Forecast hours Cases 211 178 145 118 93 49 22 20 Track error(nm) 250 200 150 100 50 0 AVNO 2014 AEMN 0 12 24 36 48 72 96 120 Forecast hours Cases 164 144 124 107 91 66 45 30
25 20 TC intensity, AL03-12, 2015 AVNO----GFS-T1534 PARA----GEFS-T574 AVNO PARA Intensity error (kts) 15 10 5 0 0 12 24 36 48 72 96 120 Forecast hours CASES 194 176 158 141 125 94 68 49